摘要
在癫痫检测任务中,脑电信号的通道选择直接影响检测性能。针对静态通道选择方法中脑电信号部分时段癫痫检测能力不足的问题,提出了动态通道选择方法。根据通道位置和脑电信号功率谱密度确定通道集合,选择通道集合中癫痫检测能力最强的一路通道作为特征提取通道,通过提高局部癫痫检测能力,进而提高整体检测能力。实验结果表明,提出的动态通道选择方法检测癫痫,取得了98.99%精确度、98.52%敏感度和99.52%特异度的较好性能。与多通道相比,检测性能相近,但特征提取通道最少,时间复杂度减少到O(1)。与单通道相比,精确度、敏感度和特异度性能指标提高4.93%以上。
In the task of epilepsy detection, the selection of EEG channel directly affects the detection performance. To solve the problem of weak detection ability in some periods of detection method using static channels, a dynamic channel selection method is proposed. The channel set is determined according to the channel position and power spectral density(PSD) of EEG. The channel with the strongest epileptic detection ability is selected as the feature extraction channel, which can enhance the overall detection ability by improving the local detection ability of epilepsy. Experimental results show that the dynamic channel selection method can detect epilepsy with 98.99% accuracy, 98.52% sensitivity and 99.52% specificity. Compared with multi-channel, the detection performance is similar. However, the feature extraction channel is the least, and the time complexity was reduced to O(1). Compared with single channel, the accuracy, sensitivity and specificity are improved more than 4.93%.
作者
汝彦冬
李金宝
吕兴凤
赵彩虹
齐景嘉
Ru Yandong;Li Jinbao;Lyu Xingfeng;Zhao Caihong;Qi Jingjia(College of Electronic Engineering,Heilongjiang University,Harbin 150080,China;College of Electronic and Information Engineering,Heilongjiang University of Science and Technology,Harbin 150027,China;Shandong Artificial Intelligence Institute,Qilu University of Technology(Shandong Academy of Science),Jinan 250014,China;College of Computer Science and Technology,Heilongjiang University,Harbin 150080,China;Teaching Affairs Department,Harbin Finance University,Harbin 150030,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2021年第2期180-188,共9页
Chinese Journal of Scientific Instrument
基金
国家重点研发计划项目(2020YFB1710200)
国家自然科学基金(61370222)
黑龙江省自然科学基金重点项目(ZD2019F003)
黑龙江省高等教育教学改革重点委托项目(SJGZ20170027)
黑龙江省属高等学校基本科研业务费基础研究项目(KJCX201815,KJCX201917)资助